Context Topic modeling finds human-readable structures in unstructured textual data. A widely used topic modeling technique is Latent Dirichlet allocation. When running on different datasets, LDA suffers from “order effects”, i.e., different topics are generated if the order of training data is shuffled. Such order effects introduce a systematic error for any study. This error can relate to misleading results; specifically, inaccurate topic descriptions and a reduction in the efficacy of text mining classification results.

Objective To provide a method in which distributions generated by LDA are more stable and can be used for further analysis.

Method We use LDADE, a search-based software engineering tool which uses Differential Evolution (DE) to tune the LDA’s parameters. LDADE is evaluated on data from a programmer information exchange site (Stackoverflow), title and abstract text of thousands of Software Engineering (SE) papers, and software defect reports from NASA. Results were collected across different implementations of LDA (Python+Scikit-Learn, Scala+Spark) across Linux platform and for different kinds of LDAs (VEM, Gibbs sampling). Results were scored via topic stability and text mining classification accuracy.

Results In all treatments: (i) standard LDA exhibits very large topic instability; (ii) LDADE’s tunings dramatically reduce cluster instability; (iii) LDADE also leads to improved performances for supervised as well as unsupervised learning.

Conclusion Due to topic instability, using standard LDA with its “off-the-shelf” settings should now be depreciated. Also, in future, we should require SE papers that use LDA to test and (if needed) mitigate LDA topic instability. Finally, LDADE is a candidate technology for effectively and efficiently reducing that instability.

Link to Publication: https://www.sciencedirect.com/science/article/pii/S0950584917300861

Amritanshu Agrawal

Wayfair

United States

Wei Fu

Department of Computer Science, North Carolina State University

United States

Tim Menzies

North Carolina State University

United States

Systems-of-systems are formed by the composition of independently created software components. These components are designed to satisfy their individual requirements, rather than the global requirements of the systems-of-systems. We refer to components that cannot be adapted to meet both individual and global requirements as defiant components. In this paper, we propose a cautious adaptation approach which supports changing the behaviour of such defiant components under exceptional conditions to satisfy global requirements, while continuing to guarantee the satisfaction of the components’ individual requirements. The approach represents both normal and exceptional conditions as scenarios; models the behaviour of exceptional conditions as wrappers implemented using an aspect-oriented technique; and deals with both single and multiple instances of defiant components with different precedence order at runtime. We evaluated an implementation of the approach using drones and boats for an organ delivery application conceived by our industrial partners, in which we assess how the proposed approach helps achieve the system-of-systems’ global requirements while accommodating increased complexity of hybrid aspects such as multiplicity, precedence ordering, openness, and heterogeneity.

Paulo Maia

State University of Ceará

Brazil

Lucas Vieira

State University of Ceará

Brazil

Matheus Chagas

State University of Ceará

Yijun Yu

The Open University, UK

United Kingdom

Andrea Zisman

The Open University

Bashar Nuseibeh

The Open University (UK) & Lero (Ireland)

United Kingdom

Ensure the correctness of safety critical systems play a key role in the worldwide software engineer. Over the past yeas we have been helping CASCO Signal Ltd which is the Chinese biggest high speed railway company to develop high speed railway safety critical software. We have also contributed specific methods for develop better safety critical software, including a search-based model-driven software development approach which uses VIATRA Solver and UML diagram refinement method to construct Sysml model and use SAT solver to check the model. This talk aims at sharing the challenge of developing high speed railway safety critical system, what we learn from develop a safety critical software with a Chinese high speed railway company, and we use ZC subsystem as a case study to show the systematic model-driven safety critical software development method.

Zhiwei Wu

East China Normal University

Jing Liu

East China Normal University

Xiang Chen

CASCO Signal Ltd.

Architecture degradation has a strong negative impact on software quality and can result in significant losses. Severe software degradation does not happen overnight. Software evolves continuously, through numerous issues, fixing bugs and adding new features, and architecture flaws emerge quietly and largely unnoticed until they grow in scope and significance when the system becomes difficult to maintain. Developers are largely unaware of these flaws or the accumulating debt as they are focused on their immediate tasks of address individual issues. As a consequence, the cumulative impacts of their activities, as they affect the architecture, go unnoticed. To detect these problems early and prevent them from accumulating into severe ones, we propose to monitor software evolution by tracking the interactions among files revised to address issues. In particular, we propose and show how we can automatically detect active hotspot, to reveal architecture problems. We have empirically studied hundreds of hotspots along the evolution timelines of 21 open source projects, and showed that there exist just a few dominating active hotspots per project at any given time. Moreover, these dominating active hotspots persist over long time periods, and thus deserve special attention. Compared with state-of-the-art design and code smell detection tools we report that by using active hotspots it is possible to detect signs of software degradation both earlier and more precisely.

Qiong Feng

Drexel University

Yuanfang Cai

Drexel University

United States

Rick Kazman

University of Hawai‘i at Mānoa

Di Cui

Xi'an Jiaotong University

Ting Liu

Xi'an Jiaotong University

China

Hongzhou Fang

Drexel University

More and more software-intensive systems employ machine learning and runtime optimization to improve their functionality by providing advanced features (e. g. personal driving assistants or recommendation engines). Such systems incorporate a number of smart software functions (SSFs) which gradually learn and adapt to the users’ preferences. A key property of SSFs is their ability to learn based on data resulting from the interaction with the user (implicit and explicit feedback)—which we call trainability. Newly developed and enhanced features in a SSF must be evaluated based on their effect on the trainability of the system. Despite recent approaches for continuous deployment of machine learning systems, trainability evaluation is not yet part of continuous integration and deployment (CID) pipelines. In this paper, we describe the different facets of trainability for the development of SSFs. We also present our approach for automated trainability evaluation within an automotive CID framework which proposes to use automated quality gates for the continuous evaluation of machine learning models. The results from our indicative evaluation based on real data from eight BMW cars highlight the importance of continuous and rigorous trainability evaluation in the development of SSFs.

Ilias Gerostathopoulos

Technical University of Munich

Germany

Stefan Kugele

Technical University of Munich

Christoph Segler

BMW Group Research, New Technologies, Innovations

Germany

Tomas Bures

Charles University, Czech Republic

Czech Republic

Alois Knoll

Technical University of Munich

Programming is typically a difficult and repetitive task. Programmers encounter endless problems during programming, and they often need to write similar code over and over again. To prevent programmers from reinventing wheels thus increase their productivity, we propose a context-aware code-to-code recommendation tool named Lancer. With the support of a Library-Sensitive Language Model (LSLM) and the BERT model, Lancer is able to automatically analyze the intention of the incomplete code and recommend relevant and reusable code samples in real-time. code, and datasets can be found. A video demonstration of Lancer can be found at https://youtu.be/tO9nhqZY35g. Lancer is open source and the code is available at https://github.com/sfzhou5678/Lancer.

Shufan Zhou

School of Electronic Information and Electrical Engineering, Shanghai Jiao Tong University

Beijun Shen

School of Electronic Information and Electrical Engineering, Shanghai Jiao Tong University

Hao Zhong

Shanghai Jiao Tong University

China